Antenatal Prediction of Shoulder Dystocia and Birth Trauma Using Routine Maternal and Ultrasound Variables

Introduction

Shoulder dystocia and birth trauma represent significant obstetric emergencies affecting 0.5-1.5% of deliveries worldwide. These complications carry substantial risks including neonatal brachial plexus injuries, hypoxic-ischemic encephalopathy, and maternal postpartum hemorrhage. Current clinical practice relies heavily on estimated fetal weight (EFW) thresholds, particularly the ≥90th centile criterion. However, EFW-based predictions show limited sensitivity. This retrospective cohort study aimed to develop superior antenatal prediction models using routinely available maternal characteristics and fetal biometric data.

Methodology

Researchers analyzed 24,334 singleton term pregnancies (≥37 weeks) delivered at a UK tertiary center between 2016-2024. All participants underwent third-trimester ultrasound at ≥36 weeks gestation. The study employed multivariable logistic regression to create two prediction models: one for shoulder dystocia (defined as requiring specific obstetric maneuvers for delivery after head emergence) and another for birth trauma (a composite outcome including shoulder dystocia, transfusion-requiring hemorrhage, full-dilatation cesarean, or neonatal hypoxic-ischemic encephalopathy). Key predictors included maternal age, BMI, parity, diabetes status, and fetal biometrics—notably abdominal circumference (AC) measured in absolute millimeters and centiles, alongside EFW measurements.

Statistical Analysis and Validation

Model performance underwent rigorous validation using bootstrapping techniques with 1,000 iterations. Researchers evaluated discrimination through area under the receiver operating characteristic curve (AUC) and assessed calibration using calibration plots and slopes. Multicollinearity between variables was systematically tested, with variance inflation factors maintained below 5. Sensitivity analyses examined performance variations across ethnic groups and diabetic subgroups. The final models were optimized using stepwise selection criteria with Akaike’s Information Criterion.

Key Findings

The AC-centile model demonstrated superior performance for both primary outcomes. For shoulder dystocia (n=432, 1.8%), the model achieved an apparent AUC of 0.706 (95% CI 0.682-0.730), with optimism-corrected validation at 0.699. For birth trauma (n=1,210, 5.0%), apparent AUC was 0.669 (95% CI 0.654-0.685), validated at 0.665. At a fixed 10% false-positive rate, the model detected 31.5% of shoulder dystocia cases and 22.8% of birth trauma cases—significantly outperforming the EFW≥90th-centile benchmark (20.4% and 14.0% sensitivity respectively). Calibration slopes indicated minimal overfitting (0.96-0.98). Maternal diabetes and fetal AC >90th centile emerged as strongest predictors.

Clinical Implications

These models represent a paradigm shift from isolated EFW thresholds toward integrated risk assessment. While discrimination remains modest (AUC<0.7 indicates limited standalone diagnostic utility), they offer valuable risk stratification for antenatal counseling. High-risk pregnancies identified through this approach could benefit from targeted interventions: delivery planning in tertiary centers, avoidance of operative vaginal delivery, and specialized staffing during labor. The findings highlight AC's crucial role in detecting asymmetric fetal growth patterns associated with shoulder dystocia.

Limitations and Future Research

The single-center design may limit generalizability, and ultrasound measurements weren’t standardized across providers. Future studies should validate these models in diverse populations and assess intervention effectiveness in high-risk cohorts identified through this methodology. Integration of placental biomarkers and machine learning approaches could further enhance prediction accuracy.

Conclusion

This research provides clinically applicable tools that outperform current EFW-based practices. Combining fetal abdominal circumference centiles with maternal risk factors enables more accurate antenatal identification of pregnancies at heightened risk for delivery complications. Though not yet suitable as standalone diagnostic tests, these models significantly advance risk-informed decision-making in obstetric care.